ViewVC Help
View File | Revision Log | Show Annotations | Download File
/cvs/JSON-XS/XS.xs
(Generate patch)

Comparing JSON-XS/XS.xs (file contents):
Revision 1.84 by root, Thu Mar 27 06:37:35 2008 UTC vs.
Revision 1.134 by root, Thu Nov 15 20:13:03 2018 UTC

6#include <string.h> 6#include <string.h>
7#include <stdlib.h> 7#include <stdlib.h>
8#include <stdio.h> 8#include <stdio.h>
9#include <limits.h> 9#include <limits.h>
10#include <float.h> 10#include <float.h>
11#include <inttypes.h>
11 12
12#if defined(__BORLANDC__) || defined(_MSC_VER) 13#if defined(__BORLANDC__) || defined(_MSC_VER)
13# define snprintf _snprintf // C compilers have this in stdio.h 14# define snprintf _snprintf // C compilers have this in stdio.h
14#endif 15#endif
15 16
16// some old perls do not have this, try to make it work, no 17// some old perls do not have this, try to make it work, no
17// guarentees, though. if it breaks, you get to keep the pieces. 18// guarantees, though. if it breaks, you get to keep the pieces.
18#ifndef UTF8_MAXBYTES 19#ifndef UTF8_MAXBYTES
19# define UTF8_MAXBYTES 13 20# define UTF8_MAXBYTES 13
20#endif 21#endif
21 22
23// compatibility with perl <5.18
24#ifndef HvNAMELEN_get
25# define HvNAMELEN_get(hv) strlen (HvNAME (hv))
26#endif
27#ifndef HvNAMELEN
28# define HvNAMELEN(hv) HvNAMELEN_get (hv)
29#endif
30#ifndef HvNAMEUTF8
31# define HvNAMEUTF8(hv) 0
32#endif
33
34// three extra for rounding, sign, and end of string
22#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 2) 35#define IVUV_MAXCHARS (sizeof (UV) * CHAR_BIT * 28 / 93 + 3)
23 36
24#define F_ASCII 0x00000001UL 37#define F_ASCII 0x00000001UL
25#define F_LATIN1 0x00000002UL 38#define F_LATIN1 0x00000002UL
26#define F_UTF8 0x00000004UL 39#define F_UTF8 0x00000004UL
27#define F_INDENT 0x00000008UL 40#define F_INDENT 0x00000008UL
32#define F_SHRINK 0x00000200UL 45#define F_SHRINK 0x00000200UL
33#define F_ALLOW_BLESSED 0x00000400UL 46#define F_ALLOW_BLESSED 0x00000400UL
34#define F_CONV_BLESSED 0x00000800UL 47#define F_CONV_BLESSED 0x00000800UL
35#define F_RELAXED 0x00001000UL 48#define F_RELAXED 0x00001000UL
36#define F_ALLOW_UNKNOWN 0x00002000UL 49#define F_ALLOW_UNKNOWN 0x00002000UL
37 50#define F_ALLOW_TAGS 0x00004000UL
38#define F_MAXDEPTH 0xf8000000UL
39#define S_MAXDEPTH 27
40#define F_MAXSIZE 0x01f00000UL
41#define S_MAXSIZE 20
42#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing 51#define F_HOOK 0x00080000UL // some hooks exist, so slow-path processing
43 52
44#define DEC_DEPTH(flags) (1UL << ((flags & F_MAXDEPTH) >> S_MAXDEPTH))
45#define DEC_SIZE(flags) (1UL << ((flags & F_MAXSIZE ) >> S_MAXSIZE ))
46
47#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER 53#define F_PRETTY F_INDENT | F_SPACE_BEFORE | F_SPACE_AFTER
48#define F_DEFAULT (9UL << S_MAXDEPTH)
49 54
50#define INIT_SIZE 32 // initial scalar size to be allocated 55#define INIT_SIZE 64 // initial scalar size to be allocated
51#define INDENT_STEP 3 // spaces per indentation level 56#define INDENT_STEP 3 // spaces per indentation level
52 57
53#define SHORT_STRING_LEN 16384 // special-case strings of up to this size 58#define SHORT_STRING_LEN 16384 // special-case strings of up to this size
59
60#define DECODE_WANTS_OCTETS(json) ((json)->flags & F_UTF8)
54 61
55#define SB do { 62#define SB do {
56#define SE } while (0) 63#define SE } while (0)
57 64
58#if __GNUC__ >= 3 65#if __GNUC__ >= 3
68 75
69#define IN_RANGE_INC(type,val,beg,end) \ 76#define IN_RANGE_INC(type,val,beg,end) \
70 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \ 77 ((unsigned type)((unsigned type)(val) - (unsigned type)(beg)) \
71 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg))) 78 <= (unsigned type)((unsigned type)(end) - (unsigned type)(beg)))
72 79
80#define ERR_NESTING_EXCEEDED "json text or perl structure exceeds maximum nesting level (max_depth set too low?)"
81
73#ifdef USE_ITHREADS 82#ifdef USE_ITHREADS
74# define JSON_SLOW 1
75# define JSON_STASH (json_stash ? json_stash : gv_stashpv ("JSON::XS", 1)) 83# define JSON_STASH (expect_true (json_stash) ? json_stash : gv_stashpv ("JSON::XS", 1))
84# define BOOL_STASH (expect_true (bool_stash) ? bool_stash : gv_stashpv ("Types::Serialiser::Boolean", 1))
85# define GET_BOOL(value) (expect_true (bool_ ## value) ? bool_ ## value : get_bool ("Types::Serialiser::" # value))
76#else 86#else
77# define JSON_SLOW 0
78# define JSON_STASH json_stash 87# define JSON_STASH json_stash
88# define BOOL_STASH bool_stash
89# define GET_BOOL(value) bool_ ## value
79#endif 90#endif
80 91
81static HV *json_stash, *json_boolean_stash; // JSON::XS:: 92// the amount of HEs to allocate on the stack, when sorting keys
82static SV *json_true, *json_false; 93#define STACK_HES 64
94
95static HV *json_stash, *bool_stash; // JSON::XS::, Types::Serialiser::Boolean::
96static SV *bool_false, *bool_true;
97static SV *sv_json;
83 98
84enum { 99enum {
85 INCR_M_WS = 0, // initial whitespace skipping, must be 0 100 INCR_M_WS = 0, // initial whitespace skipping, must be 0
86 INCR_M_STR, // inside string 101 INCR_M_STR, // inside string
87 INCR_M_BS, // inside backslash 102 INCR_M_BS, // inside backslash
103 INCR_M_C0, // inside comment in initial whitespace sequence
104 INCR_M_C1, // inside comment in other places
88 INCR_M_JSON // outside anything, count nesting 105 INCR_M_JSON // outside anything, count nesting
89}; 106};
90 107
91#define INCR_DONE(json) (!(json)->incr_nest && (json)->incr_mode == INCR_M_JSON) 108#define INCR_DONE(json) ((json)->incr_nest <= 0 && (json)->incr_mode == INCR_M_JSON)
92 109
93typedef struct { 110typedef struct {
94 U32 flags; 111 U32 flags;
112 U32 max_depth;
113 STRLEN max_size;
114
95 SV *cb_object; 115 SV *cb_object;
96 HV *cb_sk_object; 116 HV *cb_sk_object;
97 117
98 // for the incremental parser 118 // for the incremental parser
99 SV *incr_text; // the source text so far 119 SV *incr_text; // the source text so far
100 STRLEN incr_pos; // the current offset into the text 120 STRLEN incr_pos; // the current offset into the text
101 int incr_nest; // {[]}-nesting level 121 int incr_nest; // {[]}-nesting level
102 int incr_mode; 122 unsigned char incr_mode;
123
124 SV *v_false, *v_true;
103} JSON; 125} JSON;
126
127INLINE void
128json_init (JSON *json)
129{
130 Zero (json, 1, JSON);
131 json->max_depth = 512;
132}
104 133
105///////////////////////////////////////////////////////////////////////////// 134/////////////////////////////////////////////////////////////////////////////
106// utility functions 135// utility functions
107 136
108INLINE SV * 137INLINE SV *
118 147
119INLINE void 148INLINE void
120shrink (SV *sv) 149shrink (SV *sv)
121{ 150{
122 sv_utf8_downgrade (sv, 1); 151 sv_utf8_downgrade (sv, 1);
152
123 if (SvLEN (sv) > SvCUR (sv) + 1) 153 if (SvLEN (sv) > SvCUR (sv) + 1)
124 { 154 {
125#ifdef SvPV_shrink_to_cur 155#ifdef SvPV_shrink_to_cur
126 SvPV_shrink_to_cur (sv); 156 SvPV_shrink_to_cur (sv);
127#elif defined (SvPV_renew) 157#elif defined (SvPV_renew)
128 SvPV_renew (sv, SvCUR (sv) + 1); 158 SvPV_renew (sv, SvCUR (sv) + 1);
129#endif 159#endif
130 } 160 }
131} 161}
132 162
163/* adds two STRLENs together, slow, and with paranoia */
164STRLEN
165strlen_sum (STRLEN l1, STRLEN l2)
166{
167 size_t sum = l1 + l2;
168
169 if (sum < (size_t)l2 || sum != (size_t)(STRLEN)sum)
170 croak ("JSON::XS: string size overflow");
171
172 return sum;
173}
174
175/* similar to SvGROW, but somewhat safer and guarantees exponential realloc strategy */
176static char *
177json_sv_grow (SV *sv, size_t len1, size_t len2)
178{
179 len1 = strlen_sum (len1, len2);
180 len1 = strlen_sum (len1, len1 >> 1);
181
182 if (len1 > 4096 - 24)
183 len1 = (len1 | 4095) - 24;
184
185 return SvGROW (sv, len1);
186}
187
133// decode an utf-8 character and return it, or (UV)-1 in 188// decode a utf-8 character and return it, or (UV)-1 in
134// case of an error. 189// case of an error.
135// we special-case "safe" characters from U+80 .. U+7FF, 190// we special-case "safe" characters from U+80 .. U+7FF,
136// but use the very good perl function to parse anything else. 191// but use the very good perl function to parse anything else.
137// note that we never call this function for a ascii codepoints 192// note that we never call this function for a ascii codepoints
138INLINE UV 193INLINE UV
171 *s++ = 0x80 | ( ch & 0x3f); 226 *s++ = 0x80 | ( ch & 0x3f);
172 227
173 return s; 228 return s;
174} 229}
175 230
231// convert offset pointer to character index, sv must be string
232static STRLEN
233ptr_to_index (SV *sv, char *offset)
234{
235 return SvUTF8 (sv)
236 ? utf8_distance (offset, SvPVX (sv))
237 : offset - SvPVX (sv);
238}
239
240/////////////////////////////////////////////////////////////////////////////
241// fp hell
242
243// scan a group of digits, and a trailing exponent
244static void
245json_atof_scan1 (const char *s, NV *accum, int *expo, int postdp, int maxdepth)
246{
247 UV uaccum = 0;
248 int eaccum = 0;
249
250 // if we recurse too deep, skip all remaining digits
251 // to avoid a stack overflow attack
252 if (expect_false (--maxdepth <= 0))
253 while (((U8)*s - '0') < 10)
254 ++s;
255
256 for (;;)
257 {
258 U8 dig = (U8)*s - '0';
259
260 if (expect_false (dig >= 10))
261 {
262 if (dig == (U8)((U8)'.' - (U8)'0'))
263 {
264 ++s;
265 json_atof_scan1 (s, accum, expo, 1, maxdepth);
266 }
267 else if ((dig | ' ') == 'e' - '0')
268 {
269 int exp2 = 0;
270 int neg = 0;
271
272 ++s;
273
274 if (*s == '-')
275 {
276 ++s;
277 neg = 1;
278 }
279 else if (*s == '+')
280 ++s;
281
282 while ((dig = (U8)*s - '0') < 10)
283 exp2 = exp2 * 10 + *s++ - '0';
284
285 *expo += neg ? -exp2 : exp2;
286 }
287
288 break;
289 }
290
291 ++s;
292
293 uaccum = uaccum * 10 + dig;
294 ++eaccum;
295
296 // if we have too many digits, then recurse for more
297 // we actually do this for rather few digits
298 if (uaccum >= (UV_MAX - 9) / 10)
299 {
300 if (postdp) *expo -= eaccum;
301 json_atof_scan1 (s, accum, expo, postdp, maxdepth);
302 if (postdp) *expo += eaccum;
303
304 break;
305 }
306 }
307
308 // this relies greatly on the quality of the pow ()
309 // implementation of the platform, but a good
310 // implementation is hard to beat.
311 // (IEEE 754 conformant ones are required to be exact)
312 if (postdp) *expo -= eaccum;
313 *accum += uaccum * Perl_pow (10., *expo);
314 *expo += eaccum;
315}
316
317static NV
318json_atof (const char *s)
319{
320 NV accum = 0.;
321 int expo = 0;
322 int neg = 0;
323
324 if (*s == '-')
325 {
326 ++s;
327 neg = 1;
328 }
329
330 // a recursion depth of ten gives us >>500 bits
331 json_atof_scan1 (s, &accum, &expo, 0, 10);
332
333 return neg ? -accum : accum;
334}
335
336// target of scalar reference is bool? -1 == nope, 0 == false, 1 == true
337static int
338ref_bool_type (SV *sv)
339{
340 svtype svt = SvTYPE (sv);
341
342 if (svt < SVt_PVAV)
343 {
344 STRLEN len = 0;
345 char *pv = svt ? SvPV (sv, len) : 0;
346
347 if (len == 1)
348 if (*pv == '1')
349 return 1;
350 else if (*pv == '0')
351 return 0;
352 }
353
354 return -1;
355}
356
357// returns whether scalar is not a reference in the sense of allow_nonref
358static int
359json_nonref (SV *scalar)
360{
361 if (!SvROK (scalar))
362 return 1;
363
364 scalar = SvRV (scalar);
365
366 if (SvTYPE (scalar) >= SVt_PVMG)
367 {
368 if (SvSTASH (scalar) == bool_stash)
369 return 1;
370
371 if (!SvOBJECT (scalar) && ref_bool_type (scalar) >= 0)
372 return 1;
373 }
374
375 return 0;
376}
377
176///////////////////////////////////////////////////////////////////////////// 378/////////////////////////////////////////////////////////////////////////////
177// encoder 379// encoder
178 380
179// structure used for encoding JSON 381// structure used for encoding JSON
180typedef struct 382typedef struct
182 char *cur; // SvPVX (sv) + current output position 384 char *cur; // SvPVX (sv) + current output position
183 char *end; // SvEND (sv) 385 char *end; // SvEND (sv)
184 SV *sv; // result scalar 386 SV *sv; // result scalar
185 JSON json; 387 JSON json;
186 U32 indent; // indentation level 388 U32 indent; // indentation level
187 U32 maxdepth; // max. indentation/recursion level
188 UV limit; // escape character values >= this value when encoding 389 UV limit; // escape character values >= this value when encoding
189} enc_t; 390} enc_t;
190 391
191INLINE void 392INLINE void
192need (enc_t *enc, STRLEN len) 393need (enc_t *enc, STRLEN len)
193{ 394{
194 if (expect_false (enc->cur + len >= enc->end)) 395 if (expect_false ((uintptr_t)(enc->end - enc->cur) < len))
195 { 396 {
196 STRLEN cur = enc->cur - SvPVX (enc->sv); 397 STRLEN cur = enc->cur - (char *)SvPVX (enc->sv);
197 SvGROW (enc->sv, cur + len + 1); 398 char *buf = json_sv_grow (enc->sv, cur, len);
198 enc->cur = SvPVX (enc->sv) + cur; 399 enc->cur = buf + cur;
199 enc->end = SvPVX (enc->sv) + SvLEN (enc->sv) - 1; 400 enc->end = buf + SvLEN (enc->sv) - 1;
200 } 401 }
201} 402}
202 403
203INLINE void 404INLINE void
204encode_ch (enc_t *enc, char ch) 405encode_ch (enc_t *enc, char ch)
220 421
221 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case 422 if (expect_true (ch >= 0x20 && ch < 0x80)) // most common case
222 { 423 {
223 if (expect_false (ch == '"')) // but with slow exceptions 424 if (expect_false (ch == '"')) // but with slow exceptions
224 { 425 {
225 need (enc, len += 1); 426 need (enc, len + 1);
226 *enc->cur++ = '\\'; 427 *enc->cur++ = '\\';
227 *enc->cur++ = '"'; 428 *enc->cur++ = '"';
228 } 429 }
229 else if (expect_false (ch == '\\')) 430 else if (expect_false (ch == '\\'))
230 { 431 {
231 need (enc, len += 1); 432 need (enc, len + 1);
232 *enc->cur++ = '\\'; 433 *enc->cur++ = '\\';
233 *enc->cur++ = '\\'; 434 *enc->cur++ = '\\';
234 } 435 }
235 else 436 else
236 *enc->cur++ = ch; 437 *enc->cur++ = ch;
239 } 440 }
240 else 441 else
241 { 442 {
242 switch (ch) 443 switch (ch)
243 { 444 {
244 case '\010': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break; 445 case '\010': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'b'; ++str; break;
245 case '\011': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break; 446 case '\011': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 't'; ++str; break;
246 case '\012': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break; 447 case '\012': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'n'; ++str; break;
247 case '\014': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break; 448 case '\014': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'f'; ++str; break;
248 case '\015': need (enc, len += 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break; 449 case '\015': need (enc, len + 1); *enc->cur++ = '\\'; *enc->cur++ = 'r'; ++str; break;
249 450
250 default: 451 default:
251 { 452 {
252 STRLEN clen; 453 STRLEN clen;
253 UV uch; 454 UV uch;
269 if (uch >= 0x10000UL) 470 if (uch >= 0x10000UL)
270 { 471 {
271 if (uch >= 0x110000UL) 472 if (uch >= 0x110000UL)
272 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch); 473 croak ("out of range codepoint (0x%lx) encountered, unrepresentable in JSON", (unsigned long)uch);
273 474
274 need (enc, len += 11); 475 need (enc, len + 11);
275 sprintf (enc->cur, "\\u%04x\\u%04x", 476 sprintf (enc->cur, "\\u%04x\\u%04x",
276 (int)((uch - 0x10000) / 0x400 + 0xD800), 477 (int)((uch - 0x10000) / 0x400 + 0xD800),
277 (int)((uch - 0x10000) % 0x400 + 0xDC00)); 478 (int)((uch - 0x10000) % 0x400 + 0xDC00));
278 enc->cur += 12; 479 enc->cur += 12;
279 } 480 }
280 else 481 else
281 { 482 {
282 static char hexdigit [16] = "0123456789abcdef";
283 need (enc, len += 5); 483 need (enc, len + 5);
284 *enc->cur++ = '\\'; 484 *enc->cur++ = '\\';
285 *enc->cur++ = 'u'; 485 *enc->cur++ = 'u';
286 *enc->cur++ = hexdigit [ uch >> 12 ]; 486 *enc->cur++ = PL_hexdigit [ uch >> 12 ];
287 *enc->cur++ = hexdigit [(uch >> 8) & 15]; 487 *enc->cur++ = PL_hexdigit [(uch >> 8) & 15];
288 *enc->cur++ = hexdigit [(uch >> 4) & 15]; 488 *enc->cur++ = PL_hexdigit [(uch >> 4) & 15];
289 *enc->cur++ = hexdigit [(uch >> 0) & 15]; 489 *enc->cur++ = PL_hexdigit [(uch >> 0) & 15];
290 } 490 }
291 491
292 str += clen; 492 str += clen;
293 } 493 }
294 else if (enc->json.flags & F_LATIN1) 494 else if (enc->json.flags & F_LATIN1)
296 *enc->cur++ = uch; 496 *enc->cur++ = uch;
297 str += clen; 497 str += clen;
298 } 498 }
299 else if (is_utf8) 499 else if (is_utf8)
300 { 500 {
301 need (enc, len += clen); 501 need (enc, len + clen);
302 do 502 do
303 { 503 {
304 *enc->cur++ = *str++; 504 *enc->cur++ = *str++;
305 } 505 }
306 while (--clen); 506 while (--clen);
307 } 507 }
308 else 508 else
309 { 509 {
310 need (enc, len += UTF8_MAXBYTES - 1); // never more than 11 bytes needed 510 need (enc, len + UTF8_MAXBYTES - 1); // never more than 11 bytes needed
311 enc->cur = encode_utf8 (enc->cur, uch); 511 enc->cur = encode_utf8 (enc->cur, uch);
312 ++str; 512 ++str;
313 } 513 }
314 } 514 }
315 } 515 }
365static void 565static void
366encode_av (enc_t *enc, AV *av) 566encode_av (enc_t *enc, AV *av)
367{ 567{
368 int i, len = av_len (av); 568 int i, len = av_len (av);
369 569
370 if (enc->indent >= enc->maxdepth) 570 if (enc->indent >= enc->json.max_depth)
371 croak ("data structure too deep (hit recursion limit)"); 571 croak (ERR_NESTING_EXCEEDED);
372 572
373 encode_ch (enc, '['); 573 encode_ch (enc, '[');
374 574
375 if (len >= 0) 575 if (len >= 0)
376 { 576 {
377 encode_nl (enc); ++enc->indent; 577 encode_nl (enc); ++enc->indent;
378 578
379 for (i = 0; i <= len; ++i) 579 for (i = 0; i <= len; ++i)
391 encode_comma (enc); 591 encode_comma (enc);
392 } 592 }
393 593
394 encode_nl (enc); --enc->indent; encode_indent (enc); 594 encode_nl (enc); --enc->indent; encode_indent (enc);
395 } 595 }
396 596
397 encode_ch (enc, ']'); 597 encode_ch (enc, ']');
398} 598}
399 599
400static void 600static void
401encode_hk (enc_t *enc, HE *he) 601encode_hk (enc_t *enc, HE *he)
405 if (HeKLEN (he) == HEf_SVKEY) 605 if (HeKLEN (he) == HEf_SVKEY)
406 { 606 {
407 SV *sv = HeSVKEY (he); 607 SV *sv = HeSVKEY (he);
408 STRLEN len; 608 STRLEN len;
409 char *str; 609 char *str;
410 610
411 SvGETMAGIC (sv); 611 SvGETMAGIC (sv);
412 str = SvPV (sv, len); 612 str = SvPV (sv, len);
413 613
414 encode_str (enc, str, len, SvUTF8 (sv)); 614 encode_str (enc, str, len, SvUTF8 (sv));
415 } 615 }
451static void 651static void
452encode_hv (enc_t *enc, HV *hv) 652encode_hv (enc_t *enc, HV *hv)
453{ 653{
454 HE *he; 654 HE *he;
455 655
456 if (enc->indent >= enc->maxdepth) 656 if (enc->indent >= enc->json.max_depth)
457 croak ("data structure too deep (hit recursion limit)"); 657 croak (ERR_NESTING_EXCEEDED);
458 658
459 encode_ch (enc, '{'); 659 encode_ch (enc, '{');
460 660
461 // for canonical output we have to sort by keys first 661 // for canonical output we have to sort by keys first
462 // actually, this is mostly due to the stupid so-called 662 // actually, this is mostly due to the stupid so-called
463 // security workaround added somewhere in 5.8.x. 663 // security workaround added somewhere in 5.8.x
464 // that randomises hash orderings 664 // that randomises hash orderings
465 if (enc->json.flags & F_CANONICAL) 665 if (enc->json.flags & F_CANONICAL && !SvRMAGICAL (hv))
466 { 666 {
467 int count = hv_iterinit (hv); 667 int count = hv_iterinit (hv);
468 668
469 if (SvMAGICAL (hv)) 669 if (SvMAGICAL (hv))
470 { 670 {
481 } 681 }
482 682
483 if (count) 683 if (count)
484 { 684 {
485 int i, fast = 1; 685 int i, fast = 1;
486#if defined(__BORLANDC__) || defined(_MSC_VER) 686 HE *hes_stack [STACK_HES];
487 HE **hes = _alloca (count * sizeof (HE)); 687 HE **hes = hes_stack;
488#else 688
489 HE *hes [count]; // if your compiler dies here, you need to enable C99 mode 689 // allocate larger arrays on the heap
490#endif 690 if (count > STACK_HES)
691 {
692 SV *sv = sv_2mortal (NEWSV (0, count * sizeof (*hes)));
693 hes = (HE **)SvPVX (sv);
694 }
491 695
492 i = 0; 696 i = 0;
493 while ((he = hv_iternext (hv))) 697 while ((he = hv_iternext (hv)))
494 { 698 {
495 hes [i++] = he; 699 hes [i++] = he;
564// encode objects, arrays and special \0=false and \1=true values. 768// encode objects, arrays and special \0=false and \1=true values.
565static void 769static void
566encode_rv (enc_t *enc, SV *sv) 770encode_rv (enc_t *enc, SV *sv)
567{ 771{
568 svtype svt; 772 svtype svt;
773 GV *method;
569 774
570 SvGETMAGIC (sv); 775 SvGETMAGIC (sv);
571 svt = SvTYPE (sv); 776 svt = SvTYPE (sv);
572 777
573 if (expect_false (SvOBJECT (sv))) 778 if (expect_false (SvOBJECT (sv)))
574 { 779 {
575 HV *stash = !JSON_SLOW || json_boolean_stash 780 HV *stash = SvSTASH (sv);
576 ? json_boolean_stash
577 : gv_stashpv ("JSON::XS::Boolean", 1);
578 781
579 if (SvSTASH (sv) == stash) 782 if (stash == bool_stash)
580 { 783 {
581 if (SvIV (sv))
582 encode_str (enc, "true", 4, 0); 784 if (SvIV (sv)) encode_str (enc, "true" , 4, 0);
583 else
584 encode_str (enc, "false", 5, 0); 785 else encode_str (enc, "false", 5, 0);
585 } 786 }
787 else if ((enc->json.flags & F_ALLOW_TAGS) && (method = gv_fetchmethod_autoload (stash, "FREEZE", 0)))
788 {
789 int count;
790 dSP;
791
792 ENTER; SAVETMPS;
793 PUSHMARK (SP);
794 EXTEND (SP, 2);
795 // we re-bless the reference to get overload and other niceties right
796 PUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
797 PUSHs (sv_json);
798
799 PUTBACK;
800 count = call_sv ((SV *)GvCV (method), G_ARRAY);
801 SPAGAIN;
802
803 // catch this surprisingly common error
804 if (SvROK (TOPs) && SvRV (TOPs) == sv)
805 croak ("%s::FREEZE method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
806
807 encode_ch (enc, '(');
808 encode_ch (enc, '"');
809 encode_str (enc, HvNAME (stash), HvNAMELEN (stash), HvNAMEUTF8 (stash));
810 encode_ch (enc, '"');
811 encode_ch (enc, ')');
812 encode_ch (enc, '[');
813
814 if (count)
815 {
816 int i;
817
818 for (i = 0; i < count - 1; ++i)
819 {
820 encode_sv (enc, SP[i + 1 - count]);
821 encode_ch (enc, ',');
822 }
823
824 encode_sv (enc, TOPs);
825 SP -= count;
826 }
827
828 encode_ch (enc, ']');
829
830 FREETMPS; LEAVE;
831 }
832 else if ((enc->json.flags & F_CONV_BLESSED) && (method = gv_fetchmethod_autoload (stash, "TO_JSON", 0)))
833 {
834 dSP;
835
836 ENTER; SAVETMPS;
837 PUSHMARK (SP);
838 // we re-bless the reference to get overload and other niceties right
839 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), stash));
840
841 // calling with G_SCALAR ensures that we always get a 1 return value
842 PUTBACK;
843 call_sv ((SV *)GvCV (method), G_SCALAR);
844 SPAGAIN;
845
846 // catch this surprisingly common error
847 if (SvROK (TOPs) && SvRV (TOPs) == sv)
848 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
849
850 sv = POPs;
851 PUTBACK;
852
853 encode_sv (enc, sv);
854
855 FREETMPS; LEAVE;
856 }
857 else if (enc->json.flags & F_ALLOW_BLESSED)
858 encode_str (enc, "null", 4, 0);
586 else 859 else
587 {
588#if 0
589 if (0 && sv_derived_from (rv, "JSON::Literal"))
590 {
591 // not yet
592 }
593#endif
594 if (enc->json.flags & F_CONV_BLESSED)
595 {
596 // we re-bless the reference to get overload and other niceties right
597 GV *to_json = gv_fetchmethod_autoload (SvSTASH (sv), "TO_JSON", 0);
598
599 if (to_json)
600 {
601 dSP;
602
603 ENTER; SAVETMPS; PUSHMARK (SP);
604 XPUSHs (sv_bless (sv_2mortal (newRV_inc (sv)), SvSTASH (sv)));
605
606 // calling with G_SCALAR ensures that we always get a 1 return value
607 PUTBACK;
608 call_sv ((SV *)GvCV (to_json), G_SCALAR);
609 SPAGAIN;
610
611 // catch this surprisingly common error
612 if (SvROK (TOPs) && SvRV (TOPs) == sv)
613 croak ("%s::TO_JSON method returned same object as was passed instead of a new one", HvNAME (SvSTASH (sv)));
614
615 sv = POPs;
616 PUTBACK;
617
618 encode_sv (enc, sv);
619
620 FREETMPS; LEAVE;
621 }
622 else if (enc->json.flags & F_ALLOW_BLESSED)
623 encode_str (enc, "null", 4, 0);
624 else
625 croak ("encountered object '%s', but neither allow_blessed enabled nor TO_JSON method available on it",
626 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
627 }
628 else if (enc->json.flags & F_ALLOW_BLESSED)
629 encode_str (enc, "null", 4, 0);
630 else
631 croak ("encountered object '%s', but neither allow_blessed nor convert_blessed settings are enabled", 860 croak ("encountered object '%s', but neither allow_blessed, convert_blessed nor allow_tags settings are enabled (or TO_JSON/FREEZE method missing)",
632 SvPV_nolen (sv_2mortal (newRV_inc (sv)))); 861 SvPV_nolen (sv_2mortal (newRV_inc (sv))));
633 }
634 } 862 }
635 else if (svt == SVt_PVHV) 863 else if (svt == SVt_PVHV)
636 encode_hv (enc, (HV *)sv); 864 encode_hv (enc, (HV *)sv);
637 else if (svt == SVt_PVAV) 865 else if (svt == SVt_PVAV)
638 encode_av (enc, (AV *)sv); 866 encode_av (enc, (AV *)sv);
639 else if (svt < SVt_PVAV) 867 else if (svt < SVt_PVAV)
640 { 868 {
641 STRLEN len = 0; 869 int bool_type = ref_bool_type (sv);
642 char *pv = svt ? SvPV (sv, len) : 0;
643 870
644 if (len == 1 && *pv == '1') 871 if (bool_type == 1)
645 encode_str (enc, "true", 4, 0); 872 encode_str (enc, "true", 4, 0);
646 else if (len == 1 && *pv == '0') 873 else if (bool_type == 0)
647 encode_str (enc, "false", 5, 0); 874 encode_str (enc, "false", 5, 0);
648 else if (enc->json.flags & F_ALLOW_UNKNOWN) 875 else if (enc->json.flags & F_ALLOW_UNKNOWN)
649 encode_str (enc, "null", 4, 0); 876 encode_str (enc, "null", 4, 0);
650 else 877 else
651 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1", 878 croak ("cannot encode reference to scalar '%s' unless the scalar is 0 or 1",
714 } 941 }
715 else 942 else
716 { 943 {
717 // large integer, use the (rather slow) snprintf way. 944 // large integer, use the (rather slow) snprintf way.
718 need (enc, IVUV_MAXCHARS); 945 need (enc, IVUV_MAXCHARS);
719 enc->cur += 946 enc->cur +=
720 SvIsUV(sv) 947 SvIsUV(sv)
721 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv)) 948 ? snprintf (enc->cur, IVUV_MAXCHARS, "%"UVuf, (UV)SvUVX (sv))
722 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv)); 949 : snprintf (enc->cur, IVUV_MAXCHARS, "%"IVdf, (IV)SvIVX (sv));
723 } 950 }
724 } 951 }
725 else if (SvROK (sv)) 952 else if (SvROK (sv))
726 encode_rv (enc, SvRV (sv)); 953 encode_rv (enc, SvRV (sv));
727 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN) 954 else if (!SvOK (sv) || enc->json.flags & F_ALLOW_UNKNOWN)
728 encode_str (enc, "null", 4, 0); 955 encode_str (enc, "null", 4, 0);
729 else 956 else
730 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, you might want to report this", 957 croak ("encountered perl type (%s,0x%x) that JSON cannot handle, check your input data",
731 SvPV_nolen (sv), SvFLAGS (sv)); 958 SvPV_nolen (sv), (unsigned int)SvFLAGS (sv));
732} 959}
733 960
734static SV * 961static SV *
735encode_json (SV *scalar, JSON *json) 962encode_json (SV *scalar, JSON *json)
736{ 963{
737 enc_t enc; 964 enc_t enc;
738 965
739 if (!(json->flags & F_ALLOW_NONREF) && !SvROK (scalar)) 966 if (!(json->flags & F_ALLOW_NONREF) && json_nonref (scalar))
740 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)"); 967 croak ("hash- or arrayref expected (not a simple scalar, use allow_nonref to allow this)");
741 968
742 enc.json = *json; 969 enc.json = *json;
743 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE)); 970 enc.sv = sv_2mortal (NEWSV (0, INIT_SIZE));
744 enc.cur = SvPVX (enc.sv); 971 enc.cur = SvPVX (enc.sv);
745 enc.end = SvEND (enc.sv); 972 enc.end = SvEND (enc.sv);
746 enc.indent = 0; 973 enc.indent = 0;
747 enc.maxdepth = DEC_DEPTH (enc.json.flags);
748 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL 974 enc.limit = enc.json.flags & F_ASCII ? 0x000080UL
749 : enc.json.flags & F_LATIN1 ? 0x000100UL 975 : enc.json.flags & F_LATIN1 ? 0x000100UL
750 : 0x110000UL; 976 : 0x110000UL;
751 977
752 SvPOK_only (enc.sv); 978 SvPOK_only (enc.sv);
753 encode_sv (&enc, scalar); 979 encode_sv (&enc, scalar);
980 encode_nl (&enc);
754 981
755 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv)); 982 SvCUR_set (enc.sv, enc.cur - SvPVX (enc.sv));
756 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings 983 *SvEND (enc.sv) = 0; // many xs functions expect a trailing 0 for text strings
757 984
758 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8))) 985 if (!(enc.json.flags & (F_ASCII | F_LATIN1 | F_UTF8)))
819 if (*dec->cur != ch) \ 1046 if (*dec->cur != ch) \
820 ERR (# ch " expected"); \ 1047 ERR (# ch " expected"); \
821 ++dec->cur; \ 1048 ++dec->cur; \
822 SE 1049 SE
823 1050
824#define DEC_INC_DEPTH if (++dec->depth > dec->maxdepth) ERR ("json datastructure exceeds maximum nesting level (set a higher max_depth)") 1051#define DEC_INC_DEPTH if (++dec->depth > dec->json.max_depth) ERR (ERR_NESTING_EXCEEDED)
825#define DEC_DEC_DEPTH --dec->depth 1052#define DEC_DEC_DEPTH --dec->depth
826 1053
827static SV *decode_sv (dec_t *dec); 1054static SV *decode_sv (dec_t *dec);
828 1055
829static signed char decode_hexdigit[256]; 1056static signed char decode_hexdigit[256];
938 else if (expect_true (ch >= 0x20 && ch < 0x80)) 1165 else if (expect_true (ch >= 0x20 && ch < 0x80))
939 *cur++ = ch; 1166 *cur++ = ch;
940 else if (ch >= 0x80) 1167 else if (ch >= 0x80)
941 { 1168 {
942 STRLEN clen; 1169 STRLEN clen;
943 UV uch;
944 1170
945 --dec_cur; 1171 --dec_cur;
946 1172
947 uch = decode_utf8 (dec_cur, dec->end - dec_cur, &clen); 1173 decode_utf8 (dec_cur, dec->end - dec_cur, &clen);
948 if (clen == (STRLEN)-1) 1174 if (clen == (STRLEN)-1)
949 ERR ("malformed UTF-8 character in JSON string"); 1175 ERR ("malformed UTF-8 character in JSON string");
950 1176
951 do 1177 do
952 *cur++ = *dec_cur++; 1178 *cur++ = *dec_cur++;
953 while (--clen); 1179 while (--clen);
954 1180
955 utf8 = 1; 1181 utf8 = 1;
956 } 1182 }
1183 else if (ch == '\t' && dec->json.flags & F_RELAXED)
1184 *cur++ = ch;
957 else 1185 else
958 { 1186 {
959 --dec_cur; 1187 --dec_cur;
960 1188
961 if (!ch) 1189 if (!ch)
969 { 1197 {
970 STRLEN len = cur - buf; 1198 STRLEN len = cur - buf;
971 1199
972 if (sv) 1200 if (sv)
973 { 1201 {
974 SvGROW (sv, SvCUR (sv) + len + 1); 1202 STRLEN cur = SvCUR (sv);
1203
1204 if (SvLEN (sv) - cur <= len)
1205 json_sv_grow (sv, cur, len);
1206
975 memcpy (SvPVX (sv) + SvCUR (sv), buf, len); 1207 memcpy (SvPVX (sv) + SvCUR (sv), buf, len);
976 SvCUR_set (sv, SvCUR (sv) + len); 1208 SvCUR_set (sv, SvCUR (sv) + len);
977 } 1209 }
978 else 1210 else
979 sv = newSVpvn (buf, len); 1211 sv = newSVpvn (buf, len);
1070 1302
1071 // special case the rather common 1..5-digit-int case 1303 // special case the rather common 1..5-digit-int case
1072 if (*start == '-') 1304 if (*start == '-')
1073 switch (len) 1305 switch (len)
1074 { 1306 {
1075 case 2: return newSViv (-( start [1] - '0' * 1)); 1307 case 2: return newSViv (-(IV)( start [1] - '0' * 1));
1076 case 3: return newSViv (-( start [1] * 10 + start [2] - '0' * 11)); 1308 case 3: return newSViv (-(IV)( start [1] * 10 + start [2] - '0' * 11));
1077 case 4: return newSViv (-( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111)); 1309 case 4: return newSViv (-(IV)( start [1] * 100 + start [2] * 10 + start [3] - '0' * 111));
1078 case 5: return newSViv (-( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111)); 1310 case 5: return newSViv (-(IV)( start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 1111));
1079 case 6: return newSViv (-(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111)); 1311 case 6: return newSViv (-(IV)(start [1] * 10000 + start [2] * 1000 + start [3] * 100 + start [4] * 10 + start [5] - '0' * 11111));
1080 } 1312 }
1081 else 1313 else
1082 switch (len) 1314 switch (len)
1083 { 1315 {
1084 case 1: return newSViv ( start [0] - '0' * 1); 1316 case 1: return newSViv ( start [0] - '0' * 1);
1085 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11); 1317 case 2: return newSViv ( start [0] * 10 + start [1] - '0' * 11);
1086 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111); 1318 case 3: return newSViv ( start [0] * 100 + start [1] * 10 + start [2] - '0' * 111);
1087 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111); 1319 case 4: return newSViv ( start [0] * 1000 + start [1] * 100 + start [2] * 10 + start [3] - '0' * 1111);
1088 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111); 1320 case 5: return newSViv ( start [0] * 10000 + start [1] * 1000 + start [2] * 100 + start [3] * 10 + start [4] - '0' * 11111);
1089 } 1321 }
1090 1322
1091 { 1323 {
1092 UV uv; 1324 UV uv;
1093 int numtype = grok_number (start, len, &uv); 1325 int numtype = grok_number (start, len, &uv);
1102 } 1334 }
1103 1335
1104 len -= *start == '-' ? 1 : 0; 1336 len -= *start == '-' ? 1 : 0;
1105 1337
1106 // does not fit into IV or UV, try NV 1338 // does not fit into IV or UV, try NV
1107 if ((sizeof (NV) == sizeof (double) && DBL_DIG >= len) 1339 if (len <= NV_DIG)
1108 #if defined (LDBL_DIG)
1109 || (sizeof (NV) == sizeof (long double) && LDBL_DIG >= len)
1110 #endif
1111 )
1112 // fits into NV without loss of precision 1340 // fits into NV without loss of precision
1113 return newSVnv (Atof (start)); 1341 return newSVnv (json_atof (start));
1114 1342
1115 // everything else fails, convert it to a string 1343 // everything else fails, convert it to a string
1116 return newSVpvn (start, dec->cur - start); 1344 return newSVpvn (start, dec->cur - start);
1117 } 1345 }
1118 1346
1119 // loss of precision here 1347 // loss of precision here
1120 return newSVnv (Atof (start)); 1348 return newSVnv (json_atof (start));
1121 1349
1122fail: 1350fail:
1123 return 0; 1351 return 0;
1124} 1352}
1125 1353
1149 if (*dec->cur == ']') 1377 if (*dec->cur == ']')
1150 { 1378 {
1151 ++dec->cur; 1379 ++dec->cur;
1152 break; 1380 break;
1153 } 1381 }
1154 1382
1155 if (*dec->cur != ',') 1383 if (*dec->cur != ',')
1156 ERR (", or ] expected while parsing array"); 1384 ERR (", or ] expected while parsing array");
1157 1385
1158 ++dec->cur; 1386 ++dec->cur;
1159 1387
1283 1511
1284 hv_iterinit (hv); 1512 hv_iterinit (hv);
1285 he = hv_iternext (hv); 1513 he = hv_iternext (hv);
1286 hv_iterinit (hv); 1514 hv_iterinit (hv);
1287 1515
1288 // the next line creates a mortal sv each time its called. 1516 // the next line creates a mortal sv each time it's called.
1289 // might want to optimise this for common cases. 1517 // might want to optimise this for common cases.
1290 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0); 1518 cb = hv_fetch_ent (dec->json.cb_sk_object, hv_iterkeysv (he), 0, 0);
1291 1519
1292 if (cb) 1520 if (cb)
1293 { 1521 {
1294 dSP; 1522 dSP;
1295 int count; 1523 int count;
1296 1524
1297 ENTER; SAVETMPS; PUSHMARK (SP); 1525 ENTER; SAVETMPS;
1526 PUSHMARK (SP);
1298 XPUSHs (HeVAL (he)); 1527 XPUSHs (HeVAL (he));
1528 sv_2mortal (sv);
1299 1529
1300 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN; 1530 PUTBACK; count = call_sv (HeVAL (cb), G_ARRAY); SPAGAIN;
1301 1531
1302 if (count == 1) 1532 if (count == 1)
1303 { 1533 {
1304 sv = newSVsv (POPs); 1534 sv = newSVsv (POPs);
1305 FREETMPS; LEAVE; 1535 FREETMPS; LEAVE;
1306 return sv; 1536 return sv;
1307 } 1537 }
1538 else if (count)
1539 croak ("filter_json_single_key_object callbacks must not return more than one scalar");
1308 1540
1541 SvREFCNT_inc (sv);
1309 FREETMPS; LEAVE; 1542 FREETMPS; LEAVE;
1310 } 1543 }
1311 } 1544 }
1312 1545
1313 if (dec->json.cb_object) 1546 if (dec->json.cb_object)
1314 { 1547 {
1315 dSP; 1548 dSP;
1316 int count; 1549 int count;
1317 1550
1318 ENTER; SAVETMPS; PUSHMARK (SP); 1551 ENTER; SAVETMPS;
1552 PUSHMARK (SP);
1319 XPUSHs (sv_2mortal (sv)); 1553 XPUSHs (sv_2mortal (sv));
1320 1554
1321 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN; 1555 PUTBACK; count = call_sv (dec->json.cb_object, G_ARRAY); SPAGAIN;
1322 1556
1323 if (count == 1) 1557 if (count == 1)
1324 { 1558 {
1325 sv = newSVsv (POPs); 1559 sv = newSVsv (POPs);
1326 FREETMPS; LEAVE; 1560 FREETMPS; LEAVE;
1327 return sv; 1561 return sv;
1328 } 1562 }
1563 else if (count)
1564 croak ("filter_json_object callbacks must not return more than one scalar");
1329 1565
1330 SvREFCNT_inc (sv); 1566 SvREFCNT_inc (sv);
1331 FREETMPS; LEAVE; 1567 FREETMPS; LEAVE;
1332 } 1568 }
1333 } 1569 }
1339 DEC_DEC_DEPTH; 1575 DEC_DEC_DEPTH;
1340 return 0; 1576 return 0;
1341} 1577}
1342 1578
1343static SV * 1579static SV *
1580decode_tag (dec_t *dec)
1581{
1582 SV *tag = 0;
1583 SV *val = 0;
1584
1585 if (!(dec->json.flags & F_ALLOW_TAGS))
1586 ERR ("malformed JSON string, neither array, object, number, string or atom");
1587
1588 ++dec->cur;
1589
1590 decode_ws (dec);
1591
1592 tag = decode_sv (dec);
1593 if (!tag)
1594 goto fail;
1595
1596 if (!SvPOK (tag))
1597 ERR ("malformed JSON string, (tag) must be a string");
1598
1599 decode_ws (dec);
1600
1601 if (*dec->cur != ')')
1602 ERR (") expected after tag");
1603
1604 ++dec->cur;
1605
1606 decode_ws (dec);
1607
1608 val = decode_sv (dec);
1609 if (!val)
1610 goto fail;
1611
1612 if (!SvROK (val) || SvTYPE (SvRV (val)) != SVt_PVAV)
1613 ERR ("malformed JSON string, tag value must be an array");
1614
1615 {
1616 AV *av = (AV *)SvRV (val);
1617 int i, len = av_len (av) + 1;
1618 HV *stash = gv_stashsv (tag, 0);
1619 SV *sv;
1620
1621 if (!stash)
1622 ERR ("cannot decode perl-object (package does not exist)");
1623
1624 GV *method = gv_fetchmethod_autoload (stash, "THAW", 0);
1625
1626 if (!method)
1627 ERR ("cannot decode perl-object (package does not have a THAW method)");
1628
1629 dSP;
1630
1631 ENTER; SAVETMPS;
1632 PUSHMARK (SP);
1633 EXTEND (SP, len + 2);
1634 // we re-bless the reference to get overload and other niceties right
1635 PUSHs (tag);
1636 PUSHs (sv_json);
1637
1638 for (i = 0; i < len; ++i)
1639 PUSHs (*av_fetch (av, i, 1));
1640
1641 PUTBACK;
1642 call_sv ((SV *)GvCV (method), G_SCALAR);
1643 SPAGAIN;
1644
1645 SvREFCNT_dec (tag);
1646 SvREFCNT_dec (val);
1647 sv = SvREFCNT_inc (POPs);
1648
1649 PUTBACK;
1650
1651 FREETMPS; LEAVE;
1652
1653 return sv;
1654 }
1655
1656fail:
1657 SvREFCNT_dec (tag);
1658 SvREFCNT_dec (val);
1659 return 0;
1660}
1661
1662static SV *
1344decode_sv (dec_t *dec) 1663decode_sv (dec_t *dec)
1345{ 1664{
1346 // the beauty of JSON: you need exactly one character lookahead 1665 // the beauty of JSON: you need exactly one character lookahead
1347 // to parse everything. 1666 // to parse everything.
1348 switch (*dec->cur) 1667 switch (*dec->cur)
1349 { 1668 {
1350 case '"': ++dec->cur; return decode_str (dec); 1669 case '"': ++dec->cur; return decode_str (dec);
1351 case '[': ++dec->cur; return decode_av (dec); 1670 case '[': ++dec->cur; return decode_av (dec);
1352 case '{': ++dec->cur; return decode_hv (dec); 1671 case '{': ++dec->cur; return decode_hv (dec);
1672 case '(': return decode_tag (dec);
1353 1673
1354 case '-': 1674 case '-':
1355 case '0': case '1': case '2': case '3': case '4': 1675 case '0': case '1': case '2': case '3': case '4':
1356 case '5': case '6': case '7': case '8': case '9': 1676 case '5': case '6': case '7': case '8': case '9':
1357 return decode_num (dec); 1677 return decode_num (dec);
1358 1678
1679 case 'f':
1680 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1681 {
1682 dec->cur += 5;
1683
1684 if (expect_false (!dec->json.v_false))
1685 dec->json.v_false = GET_BOOL (false);
1686
1687 return newSVsv (dec->json.v_false);
1688 }
1689 else
1690 ERR ("'false' expected");
1691
1692 break;
1693
1359 case 't': 1694 case 't':
1360 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4)) 1695 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "true", 4))
1361 { 1696 {
1362 dec->cur += 4; 1697 dec->cur += 4;
1363#if JSON_SLOW 1698
1364 json_true = get_bool ("JSON::XS::true"); 1699 if (expect_false (!dec->json.v_true))
1365#endif 1700 dec->json.v_true = GET_BOOL (true);
1701
1366 return newSVsv (json_true); 1702 return newSVsv (dec->json.v_true);
1367 } 1703 }
1368 else 1704 else
1369 ERR ("'true' expected"); 1705 ERR ("'true' expected");
1370
1371 break;
1372
1373 case 'f':
1374 if (dec->end - dec->cur >= 5 && !memcmp (dec->cur, "false", 5))
1375 {
1376 dec->cur += 5;
1377#if JSON_SLOW
1378 json_false = get_bool ("JSON::XS::false");
1379#endif
1380 return newSVsv (json_false);
1381 }
1382 else
1383 ERR ("'false' expected");
1384 1706
1385 break; 1707 break;
1386 1708
1387 case 'n': 1709 case 'n':
1388 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4)) 1710 if (dec->end - dec->cur >= 4 && !memcmp (dec->cur, "null", 4))
1394 ERR ("'null' expected"); 1716 ERR ("'null' expected");
1395 1717
1396 break; 1718 break;
1397 1719
1398 default: 1720 default:
1399 ERR ("malformed JSON string, neither array, object, number, string or atom"); 1721 ERR ("malformed JSON string, neither tag, array, object, number, string or atom");
1400 break; 1722 break;
1401 } 1723 }
1402 1724
1403fail: 1725fail:
1404 return 0; 1726 return 0;
1406 1728
1407static SV * 1729static SV *
1408decode_json (SV *string, JSON *json, STRLEN *offset_return) 1730decode_json (SV *string, JSON *json, STRLEN *offset_return)
1409{ 1731{
1410 dec_t dec; 1732 dec_t dec;
1411 STRLEN offset;
1412 SV *sv; 1733 SV *sv;
1413 1734
1735 /* work around bugs in 5.10 where manipulating magic values
1736 * makes perl ignore the magic in subsequent accesses.
1737 * also make a copy of non-PV values, to get them into a clean
1738 * state (SvPV should do that, but it's buggy, see below).
1739 *
1740 * SvIsCOW_shared_hash works around a bug in perl (possibly 5.16),
1741 * as reported by Reini Urban.
1742 */
1414 SvGETMAGIC (string); 1743 /*SvGETMAGIC (string);*/
1744 if (SvMAGICAL (string) || !SvPOK (string) || SvIsCOW_shared_hash (string))
1745 string = sv_2mortal (newSVsv (string));
1746
1415 SvUPGRADE (string, SVt_PV); 1747 SvUPGRADE (string, SVt_PV);
1416 1748
1417 if (json->flags & F_MAXSIZE && SvCUR (string) > DEC_SIZE (json->flags)) 1749 /* work around a bug in perl 5.10, which causes SvCUR to fail an
1750 * assertion with -DDEBUGGING, although SvCUR is documented to
1751 * return the xpv_cur field which certainly exists after upgrading.
1752 * according to nicholas clark, calling SvPOK fixes this.
1753 * But it doesn't fix it, so try another workaround, call SvPV_nolen
1754 * and hope for the best.
1755 * Damnit, SvPV_nolen still trips over yet another assertion. This
1756 * assertion business is seriously broken, try yet another workaround
1757 * for the broken -DDEBUGGING.
1758 */
1759 {
1760#ifdef DEBUGGING
1761 STRLEN offset = SvOK (string) ? sv_len (string) : 0;
1762#else
1763 STRLEN offset = SvCUR (string);
1764#endif
1765
1766 if (offset > json->max_size && json->max_size)
1418 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu", 1767 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
1419 (unsigned long)SvCUR (string), (unsigned long)DEC_SIZE (json->flags)); 1768 (unsigned long)SvCUR (string), (unsigned long)json->max_size);
1769 }
1420 1770
1421 if (json->flags & F_UTF8) 1771 if (DECODE_WANTS_OCTETS (json))
1422 sv_utf8_downgrade (string, 0); 1772 sv_utf8_downgrade (string, 0);
1423 else 1773 else
1424 sv_utf8_upgrade (string); 1774 sv_utf8_upgrade (string);
1425 1775
1426 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP 1776 SvGROW (string, SvCUR (string) + 1); // should basically be a NOP
1427 1777
1428 dec.json = *json; 1778 dec.json = *json;
1429 dec.cur = SvPVX (string); 1779 dec.cur = SvPVX (string);
1430 dec.end = SvEND (string); 1780 dec.end = SvEND (string);
1431 dec.err = 0; 1781 dec.err = 0;
1432 dec.depth = 0; 1782 dec.depth = 0;
1433 dec.maxdepth = DEC_DEPTH (dec.json.flags);
1434 1783
1435 if (dec.json.cb_object || dec.json.cb_sk_object) 1784 if (dec.json.cb_object || dec.json.cb_sk_object)
1436 dec.json.flags |= F_HOOK; 1785 dec.json.flags |= F_HOOK;
1437 1786
1438 *dec.end = 0; // this should basically be a nop, too, but make sure it's there 1787 *dec.end = 0; // this should basically be a nop, too, but make sure it's there
1439 1788
1440 decode_ws (&dec); 1789 decode_ws (&dec);
1441 sv = decode_sv (&dec); 1790 sv = decode_sv (&dec);
1442 1791
1443 if (!(offset_return || !sv)) 1792 if (offset_return)
1793 *offset_return = dec.cur - SvPVX (string);
1794 else if (sv)
1444 { 1795 {
1445 // check for trailing garbage 1796 // check for trailing garbage
1446 decode_ws (&dec); 1797 decode_ws (&dec);
1447 1798
1448 if (*dec.cur) 1799 if (*dec.cur)
1449 { 1800 {
1450 dec.err = "garbage after JSON object"; 1801 dec.err = "garbage after JSON object";
1451 SvREFCNT_dec (sv); 1802 SvREFCNT_dec (sv);
1452 sv = 0; 1803 sv = 0;
1453 } 1804 }
1454 }
1455
1456 if (offset_return || !sv)
1457 {
1458 offset = dec.json.flags & F_UTF8
1459 ? dec.cur - SvPVX (string)
1460 : utf8_distance (dec.cur, SvPVX (string));
1461
1462 if (offset_return)
1463 *offset_return = offset;
1464 } 1805 }
1465 1806
1466 if (!sv) 1807 if (!sv)
1467 { 1808 {
1468 SV *uni = sv_newmortal (); 1809 SV *uni = sv_newmortal ();
1474 SAVEVPTR (PL_curcop); 1815 SAVEVPTR (PL_curcop);
1475 PL_curcop = &cop; 1816 PL_curcop = &cop;
1476 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ); 1817 pv_uni_display (uni, dec.cur, dec.end - dec.cur, 20, UNI_DISPLAY_QQ);
1477 LEAVE; 1818 LEAVE;
1478 1819
1479 croak ("%s, at character offset %d [\"%s\"]", 1820 croak ("%s, at character offset %d (before \"%s\")",
1480 dec.err, 1821 dec.err,
1481 (int)offset, 1822 (int)ptr_to_index (string, dec.cur),
1482 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)"); 1823 dec.cur != dec.end ? SvPV_nolen (uni) : "(end of string)");
1483 } 1824 }
1484 1825
1485 sv = sv_2mortal (sv); 1826 sv = sv_2mortal (sv);
1486 1827
1487 if (!(dec.json.flags & F_ALLOW_NONREF) && !SvROK (sv)) 1828 if (!(dec.json.flags & F_ALLOW_NONREF) && json_nonref (sv))
1488 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)"); 1829 croak ("JSON text must be an object or array (but found number, string, true, false or null, use allow_nonref to allow this)");
1489 1830
1490 return sv; 1831 return sv;
1491} 1832}
1492 1833
1495 1836
1496static void 1837static void
1497incr_parse (JSON *self) 1838incr_parse (JSON *self)
1498{ 1839{
1499 const char *p = SvPVX (self->incr_text) + self->incr_pos; 1840 const char *p = SvPVX (self->incr_text) + self->incr_pos;
1841
1842 // the state machine here is a bit convoluted and could be simplified a lot
1843 // but this would make it slower, so...
1500 1844
1501 for (;;) 1845 for (;;)
1502 { 1846 {
1503 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D 1847 //printf ("loop pod %d *p<%c><%s>, mode %d nest %d\n", p - SvPVX (self->incr_text), *p, p, self->incr_mode, self->incr_nest);//D
1504 switch (self->incr_mode) 1848 switch (self->incr_mode)
1505 { 1849 {
1506 // only used for intiial whitespace skipping 1850 // only used for initial whitespace skipping
1507 case INCR_M_WS: 1851 case INCR_M_WS:
1508 for (;;) 1852 for (;;)
1509 { 1853 {
1510 if (*p > 0x20) 1854 if (*p > 0x20)
1511 { 1855 {
1856 if (*p == '#')
1857 {
1858 self->incr_mode = INCR_M_C0;
1859 goto incr_m_c;
1860 }
1861 else
1862 {
1512 self->incr_mode = INCR_M_JSON; 1863 self->incr_mode = INCR_M_JSON;
1513 goto incr_m_json; 1864 goto incr_m_json;
1865 }
1514 } 1866 }
1515 else if (!*p) 1867 else if (!*p)
1516 goto interrupt; 1868 goto interrupt;
1517 1869
1518 ++p; 1870 ++p;
1524 goto interrupt; 1876 goto interrupt;
1525 1877
1526 ++p; 1878 ++p;
1527 self->incr_mode = INCR_M_STR; 1879 self->incr_mode = INCR_M_STR;
1528 goto incr_m_str; 1880 goto incr_m_str;
1881
1882 // inside #-style comments
1883 case INCR_M_C0:
1884 case INCR_M_C1:
1885 incr_m_c:
1886 for (;;)
1887 {
1888 if (*p == '\n')
1889 {
1890 self->incr_mode = self->incr_mode == INCR_M_C0 ? INCR_M_WS : INCR_M_JSON;
1891 break;
1892 }
1893 else if (!*p)
1894 goto interrupt;
1895
1896 ++p;
1897 }
1898
1899 break;
1529 1900
1530 // inside a string 1901 // inside a string
1531 case INCR_M_STR: 1902 case INCR_M_STR:
1532 incr_m_str: 1903 incr_m_str:
1533 for (;;) 1904 for (;;)
1584 self->incr_mode = INCR_M_STR; 1955 self->incr_mode = INCR_M_STR;
1585 goto incr_m_str; 1956 goto incr_m_str;
1586 1957
1587 case '[': 1958 case '[':
1588 case '{': 1959 case '{':
1960 case '(':
1589 ++self->incr_nest; 1961 if (++self->incr_nest > self->max_depth)
1962 croak (ERR_NESTING_EXCEEDED);
1590 break; 1963 break;
1591 1964
1592 case ']': 1965 case ']':
1593 case '}': 1966 case '}':
1594 if (!--self->incr_nest) 1967 if (--self->incr_nest <= 0)
1595 goto interrupt; 1968 goto interrupt;
1969 break;
1970
1971 case ')':
1972 --self->incr_nest;
1973 break;
1974
1975 case '#':
1976 self->incr_mode = INCR_M_C1;
1977 goto incr_m_c;
1596 } 1978 }
1597 } 1979 }
1598 } 1980 }
1599 1981
1600 modechange: 1982 modechange:
1601 ; 1983 ;
1602 } 1984 }
1603 1985
1604interrupt: 1986interrupt:
1605 self->incr_pos = p - SvPVX (self->incr_text); 1987 self->incr_pos = p - SvPVX (self->incr_text);
1988 //printf ("interrupt<%.*s>\n", self->incr_pos, SvPVX(self->incr_text));//D
1606 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D 1989 //printf ("return pos %d mode %d nest %d\n", self->incr_pos, self->incr_mode, self->incr_nest);//D
1607} 1990}
1608 1991
1609///////////////////////////////////////////////////////////////////////////// 1992/////////////////////////////////////////////////////////////////////////////
1610// XS interface functions 1993// XS interface functions
1620 i >= '0' && i <= '9' ? i - '0' 2003 i >= '0' && i <= '9' ? i - '0'
1621 : i >= 'a' && i <= 'f' ? i - 'a' + 10 2004 : i >= 'a' && i <= 'f' ? i - 'a' + 10
1622 : i >= 'A' && i <= 'F' ? i - 'A' + 10 2005 : i >= 'A' && i <= 'F' ? i - 'A' + 10
1623 : -1; 2006 : -1;
1624 2007
1625 json_stash = gv_stashpv ("JSON::XS" , 1); 2008 json_stash = gv_stashpv ("JSON::XS" , 1);
1626 json_boolean_stash = gv_stashpv ("JSON::XS::Boolean", 1); 2009 bool_stash = gv_stashpv ("Types::Serialiser::Boolean", 1);
1627
1628 json_true = get_bool ("JSON::XS::true");
1629 json_false = get_bool ("JSON::XS::false"); 2010 bool_false = get_bool ("Types::Serialiser::false");
2011 bool_true = get_bool ("Types::Serialiser::true");
2012
2013 sv_json = newSVpv ("JSON", 0);
2014 SvREADONLY_on (sv_json);
2015
2016 CvNODEBUG_on (get_cv ("JSON::XS::incr_text", 0)); /* the debugger completely breaks lvalue subs */
1630} 2017}
1631 2018
1632PROTOTYPES: DISABLE 2019PROTOTYPES: DISABLE
1633 2020
1634void CLONE (...) 2021void CLONE (...)
1635 CODE: 2022 CODE:
2023 // as long as these writes are atomic, the race should not matter
2024 // as existing threads either already use 0, or use the old value,
2025 // which is sitll correct for the initial thread.
1636 json_stash = 0; 2026 json_stash = 0;
1637 json_boolean_stash = 0; 2027 bool_stash = 0;
2028 bool_false = 0;
2029 bool_true = 0;
1638 2030
1639void new (char *klass) 2031void new (char *klass)
1640 PPCODE: 2032 PPCODE:
1641{ 2033{
1642 SV *pv = NEWSV (0, sizeof (JSON)); 2034 SV *pv = NEWSV (0, sizeof (JSON));
1643 SvPOK_only (pv); 2035 SvPOK_only (pv);
1644 Zero (SvPVX (pv), 1, JSON); 2036 json_init ((JSON *)SvPVX (pv));
1645 ((JSON *)SvPVX (pv))->flags = F_DEFAULT;
1646 XPUSHs (sv_2mortal (sv_bless ( 2037 XPUSHs (sv_2mortal (sv_bless (
1647 newRV_noinc (pv), 2038 newRV_noinc (pv),
1648 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1) 2039 strEQ (klass, "JSON::XS") ? JSON_STASH : gv_stashpv (klass, 1)
1649 ))); 2040 )));
1650} 2041}
2042
2043void boolean_values (JSON *self, SV *v_false = 0, SV *v_true = 0)
2044 PPCODE:
2045 self->v_false = newSVsv (v_false);
2046 self->v_true = newSVsv (v_true);
2047 XPUSHs (ST (0));
2048
2049void get_boolean_values (JSON *self)
2050 PPCODE:
2051 if (self->v_false && self->v_true)
2052 {
2053 EXTEND (SP, 2);
2054 PUSHs (self->v_false);
2055 PUSHs (self->v_true);
2056 }
1651 2057
1652void ascii (JSON *self, int enable = 1) 2058void ascii (JSON *self, int enable = 1)
1653 ALIAS: 2059 ALIAS:
1654 ascii = F_ASCII 2060 ascii = F_ASCII
1655 latin1 = F_LATIN1 2061 latin1 = F_LATIN1
1663 shrink = F_SHRINK 2069 shrink = F_SHRINK
1664 allow_blessed = F_ALLOW_BLESSED 2070 allow_blessed = F_ALLOW_BLESSED
1665 convert_blessed = F_CONV_BLESSED 2071 convert_blessed = F_CONV_BLESSED
1666 relaxed = F_RELAXED 2072 relaxed = F_RELAXED
1667 allow_unknown = F_ALLOW_UNKNOWN 2073 allow_unknown = F_ALLOW_UNKNOWN
2074 allow_tags = F_ALLOW_TAGS
1668 PPCODE: 2075 PPCODE:
1669{ 2076{
1670 if (enable) 2077 if (enable)
1671 self->flags |= ix; 2078 self->flags |= ix;
1672 else 2079 else
1688 get_shrink = F_SHRINK 2095 get_shrink = F_SHRINK
1689 get_allow_blessed = F_ALLOW_BLESSED 2096 get_allow_blessed = F_ALLOW_BLESSED
1690 get_convert_blessed = F_CONV_BLESSED 2097 get_convert_blessed = F_CONV_BLESSED
1691 get_relaxed = F_RELAXED 2098 get_relaxed = F_RELAXED
1692 get_allow_unknown = F_ALLOW_UNKNOWN 2099 get_allow_unknown = F_ALLOW_UNKNOWN
2100 get_allow_tags = F_ALLOW_TAGS
1693 PPCODE: 2101 PPCODE:
1694 XPUSHs (boolSV (self->flags & ix)); 2102 XPUSHs (boolSV (self->flags & ix));
1695 2103
1696void max_depth (JSON *self, UV max_depth = 0x80000000UL) 2104void max_depth (JSON *self, U32 max_depth = 0x80000000UL)
1697 PPCODE: 2105 PPCODE:
1698{ 2106 self->max_depth = max_depth;
1699 UV log2 = 0;
1700
1701 if (max_depth > 0x80000000UL) max_depth = 0x80000000UL;
1702
1703 while ((1UL << log2) < max_depth)
1704 ++log2;
1705
1706 self->flags = self->flags & ~F_MAXDEPTH | (log2 << S_MAXDEPTH);
1707
1708 XPUSHs (ST (0)); 2107 XPUSHs (ST (0));
1709}
1710 2108
1711U32 get_max_depth (JSON *self) 2109U32 get_max_depth (JSON *self)
1712 CODE: 2110 CODE:
1713 RETVAL = DEC_DEPTH (self->flags); 2111 RETVAL = self->max_depth;
1714 OUTPUT: 2112 OUTPUT:
1715 RETVAL 2113 RETVAL
1716 2114
1717void max_size (JSON *self, UV max_size = 0) 2115void max_size (JSON *self, U32 max_size = 0)
1718 PPCODE: 2116 PPCODE:
1719{ 2117 self->max_size = max_size;
1720 UV log2 = 0;
1721
1722 if (max_size > 0x80000000UL) max_size = 0x80000000UL;
1723 if (max_size == 1) max_size = 2;
1724
1725 while ((1UL << log2) < max_size)
1726 ++log2;
1727
1728 self->flags = self->flags & ~F_MAXSIZE | (log2 << S_MAXSIZE);
1729
1730 XPUSHs (ST (0)); 2118 XPUSHs (ST (0));
1731}
1732 2119
1733int get_max_size (JSON *self) 2120int get_max_size (JSON *self)
1734 CODE: 2121 CODE:
1735 RETVAL = DEC_SIZE (self->flags); 2122 RETVAL = self->max_size;
1736 OUTPUT: 2123 OUTPUT:
1737 RETVAL 2124 RETVAL
1738 2125
1739void filter_json_object (JSON *self, SV *cb = &PL_sv_undef) 2126void filter_json_object (JSON *self, SV *cb = &PL_sv_undef)
1740 PPCODE: 2127 PPCODE:
1746} 2133}
1747 2134
1748void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef) 2135void filter_json_single_key_object (JSON *self, SV *key, SV *cb = &PL_sv_undef)
1749 PPCODE: 2136 PPCODE:
1750{ 2137{
1751 if (!self->cb_sk_object) 2138 if (!self->cb_sk_object)
1752 self->cb_sk_object = newHV (); 2139 self->cb_sk_object = newHV ();
1753 2140
1754 if (SvOK (cb)) 2141 if (SvOK (cb))
1755 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0); 2142 hv_store_ent (self->cb_sk_object, key, newSVsv (cb), 0);
1756 else 2143 else
1767 XPUSHs (ST (0)); 2154 XPUSHs (ST (0));
1768} 2155}
1769 2156
1770void encode (JSON *self, SV *scalar) 2157void encode (JSON *self, SV *scalar)
1771 PPCODE: 2158 PPCODE:
1772 XPUSHs (encode_json (scalar, self)); 2159 PUTBACK; scalar = encode_json (scalar, self); SPAGAIN;
2160 XPUSHs (scalar);
1773 2161
1774void decode (JSON *self, SV *jsonstr) 2162void decode (JSON *self, SV *jsonstr)
1775 PPCODE: 2163 PPCODE:
1776 XPUSHs (decode_json (jsonstr, self, 0)); 2164 PUTBACK; jsonstr = decode_json (jsonstr, self, 0); SPAGAIN;
2165 XPUSHs (jsonstr);
1777 2166
1778void decode_prefix (JSON *self, SV *jsonstr) 2167void decode_prefix (JSON *self, SV *jsonstr)
1779 PPCODE: 2168 PPCODE:
1780{ 2169{
2170 SV *sv;
1781 STRLEN offset; 2171 STRLEN offset;
2172 PUTBACK; sv = decode_json (jsonstr, self, &offset); SPAGAIN;
1782 EXTEND (SP, 2); 2173 EXTEND (SP, 2);
1783 PUSHs (decode_json (jsonstr, self, &offset)); 2174 PUSHs (sv);
1784 PUSHs (sv_2mortal (newSVuv (offset))); 2175 PUSHs (sv_2mortal (newSVuv (ptr_to_index (jsonstr, SvPV_nolen (jsonstr) + offset))));
1785} 2176}
1786 2177
1787void incr_parse (JSON *self, SV *jsonstr = 0) 2178void incr_parse (JSON *self, SV *jsonstr = 0)
1788 PPCODE: 2179 PPCODE:
1789{ 2180{
1790 if (!self->incr_text) 2181 if (!self->incr_text)
1791 self->incr_text = newSVpvn ("", 0); 2182 self->incr_text = newSVpvn ("", 0);
2183
2184 /* if utf8-ness doesn't match the decoder, need to upgrade/downgrade */
2185 if (!DECODE_WANTS_OCTETS (self) == !SvUTF8 (self->incr_text))
2186 if (DECODE_WANTS_OCTETS (self))
2187 {
2188 if (self->incr_pos)
2189 self->incr_pos = utf8_length ((U8 *)SvPVX (self->incr_text),
2190 (U8 *)SvPVX (self->incr_text) + self->incr_pos);
2191
2192 sv_utf8_downgrade (self->incr_text, 0);
2193 }
2194 else
2195 {
2196 sv_utf8_upgrade (self->incr_text);
2197
2198 if (self->incr_pos)
2199 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
2200 - (U8 *)SvPVX (self->incr_text);
2201 }
1792 2202
1793 // append data, if any 2203 // append data, if any
1794 if (jsonstr) 2204 if (jsonstr)
1795 { 2205 {
2206 /* make sure both strings have same encoding */
1796 if (SvUTF8 (jsonstr) && !SvUTF8 (self->incr_text)) 2207 if (SvUTF8 (jsonstr) != SvUTF8 (self->incr_text))
2208 if (SvUTF8 (jsonstr))
2209 sv_utf8_downgrade (jsonstr, 0);
1797 { 2210 else
1798 /* utf-8-ness differs, need to upgrade */
1799 sv_utf8_upgrade (self->incr_text); 2211 sv_utf8_upgrade (jsonstr);
1800 2212
1801 if (self->incr_pos) 2213 /* and then just blindly append */
1802 self->incr_pos = utf8_hop ((U8 *)SvPVX (self->incr_text), self->incr_pos)
1803 - (U8 *)SvPVX (self->incr_text);
1804 }
1805
1806 { 2214 {
1807 STRLEN len; 2215 STRLEN len;
1808 const char *str = SvPV (jsonstr, len); 2216 const char *str = SvPV (jsonstr, len);
1809 SvGROW (self->incr_text, SvCUR (self->incr_text) + len + 1); 2217 STRLEN cur = SvCUR (self->incr_text);
2218
2219 if (SvLEN (self->incr_text) - cur <= len)
2220 json_sv_grow (self->incr_text, cur, len);
2221
1810 Move (str, SvEND (self->incr_text), len, char); 2222 Move (str, SvEND (self->incr_text), len, char);
1811 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len); 2223 SvCUR_set (self->incr_text, SvCUR (self->incr_text) + len);
1812 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there 2224 *SvEND (self->incr_text) = 0; // this should basically be a nop, too, but make sure it's there
1813 } 2225 }
1814 } 2226 }
1815 2227
1816 if (GIMME_V != G_VOID) 2228 if (GIMME_V != G_VOID)
1817 do 2229 do
1818 { 2230 {
2231 SV *sv;
1819 STRLEN offset; 2232 STRLEN offset;
1820 2233
1821 if (!INCR_DONE (self)) 2234 if (!INCR_DONE (self))
1822 { 2235 {
1823 incr_parse (self); 2236 incr_parse (self);
2237
2238 if (self->incr_pos > self->max_size && self->max_size)
2239 croak ("attempted decode of JSON text of %lu bytes size, but max_size is set to %lu",
2240 (unsigned long)self->incr_pos, (unsigned long)self->max_size);
2241
1824 if (!INCR_DONE (self)) 2242 if (!INCR_DONE (self))
2243 {
2244 // as an optimisation, do not accumulate white space in the incr buffer
2245 if (self->incr_mode == INCR_M_WS && self->incr_pos)
2246 {
2247 self->incr_pos = 0;
2248 SvCUR_set (self->incr_text, 0);
2249 }
2250
1825 break; 2251 break;
2252 }
1826 } 2253 }
1827 2254
1828 XPUSHs (decode_json (self->incr_text, self, &offset)); 2255 PUTBACK; sv = decode_json (self->incr_text, self, &offset); SPAGAIN;
2256 XPUSHs (sv);
1829 2257
1830 sv_chop (self->incr_text, SvPV_nolen (self->incr_text) + offset);
1831 self->incr_pos -= offset; 2258 self->incr_pos -= offset;
1832 self->incr_nest = 0; 2259 self->incr_nest = 0;
1833 self->incr_mode = 0; 2260 self->incr_mode = 0;
2261
2262 sv_chop (self->incr_text, SvPVX (self->incr_text) + offset);
1834 } 2263 }
1835 while (GIMME_V == G_ARRAY); 2264 while (GIMME_V == G_ARRAY);
1836} 2265}
1837 2266
1838SV *incr_text (JSON *self) 2267SV *incr_text (JSON *self)
1857 self->incr_nest = 0; 2286 self->incr_nest = 0;
1858 self->incr_mode = 0; 2287 self->incr_mode = 0;
1859 } 2288 }
1860} 2289}
1861 2290
2291void incr_reset (JSON *self)
2292 CODE:
2293{
2294 SvREFCNT_dec (self->incr_text);
2295 self->incr_text = 0;
2296 self->incr_pos = 0;
2297 self->incr_nest = 0;
2298 self->incr_mode = 0;
2299}
2300
1862void DESTROY (JSON *self) 2301void DESTROY (JSON *self)
1863 CODE: 2302 CODE:
2303 SvREFCNT_dec (self->v_false);
2304 SvREFCNT_dec (self->v_true);
1864 SvREFCNT_dec (self->cb_sk_object); 2305 SvREFCNT_dec (self->cb_sk_object);
1865 SvREFCNT_dec (self->cb_object); 2306 SvREFCNT_dec (self->cb_object);
1866 SvREFCNT_dec (self->incr_text); 2307 SvREFCNT_dec (self->incr_text);
1867 2308
1868PROTOTYPES: ENABLE 2309PROTOTYPES: ENABLE
1869 2310
1870void encode_json (SV *scalar) 2311void encode_json (SV *scalar)
1871 ALIAS:
1872 to_json_ = 0
1873 encode_json = F_UTF8
1874 PPCODE: 2312 PPCODE:
1875{ 2313{
1876 JSON json = { F_DEFAULT | ix }; 2314 JSON json;
1877 XPUSHs (encode_json (scalar, &json)); 2315 json_init (&json);
2316 json.flags |= F_UTF8;
2317 PUTBACK; scalar = encode_json (scalar, &json); SPAGAIN;
2318 XPUSHs (scalar);
1878} 2319}
1879 2320
1880void decode_json (SV *jsonstr) 2321void decode_json (SV *jsonstr)
1881 ALIAS:
1882 from_json_ = 0
1883 decode_json = F_UTF8
1884 PPCODE: 2322 PPCODE:
1885{ 2323{
1886 JSON json = { F_DEFAULT | ix }; 2324 JSON json;
1887 XPUSHs (decode_json (jsonstr, &json, 0)); 2325 json_init (&json);
2326 json.flags |= F_UTF8;
2327 PUTBACK; jsonstr = decode_json (jsonstr, &json, 0); SPAGAIN;
2328 XPUSHs (jsonstr);
1888} 2329}
1889 2330
1890

Diff Legend

Removed lines
+ Added lines
< Changed lines
> Changed lines